plots in Figures 3 and 4 show relatively high output power in
this region. Numerical results show that the output power
Figure 3 Tuning characteristics of Er 3q-doped ber ring laser.
Inset: a numerical spectral plot tuned at 1515 nm. Notice a small
hump in the 1530 nm region having a similar prole as in Figure 4
arrow.
Tuning the laser to a wavelength shorter than 1535 nm
brings about a signicant drop in the output power level; see
Figure 3. The drop in power level is due to higher intracavity
losses incurred by the variation of the insertion loss of the
tunable lter. The plot shows the continuous tuning charac-
teristic of the laser, with small power variations in the
15351567 nm tuning range. Under certain conditions, with
regard to its intracavity loss, output coupler ratio, and EDF
length, an EDFL without a lter in the cavity. will lase in
either the 1530 or 1550 nm region, but not in the 1540 nm
region. However, the laser with a lter incorporated in the
cavity can be tuned and lase at 1540 nm. A numerical study
shows that the lter functions such that only modes within
the lter bandwidth are allowed to oscillate and are ampli-
ed; hence, all of the power is pumped and channeled into
the tuned bandwidth. A numerical spectral plot of the laser
output tuned at 1515 nm is shown in the inset of Figure 3. A
small hump in the 1530 nm region, which illustrates a higher
amplied spontaneous emission ASE. level that results from
high absorption and gain coefcient at 1530 nm, demon-
strates a similar prole arrow. in Figure 4. The numerical
plot shows a close match with the measured result. The
measured output spectra of the EDFL tuned from 1515 to
1567 nm are shown in Figure 4, with a signal-to-noise ratio
SNR. of about 70 dB. The tuning limitation of the lter
restrains further measurement beyond 1567 nm, although the
Figure 4 EDFL output spectra when tuned from 1515 to 1567 nm.
The 3 dB spectral width is 0.044 nm, limited by the optical spectrum
analyzer resolution bandwidth
MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 29, No. 4, May 20 2001
drops substantially when the laser is tuned to 1570 nm and at
a longer wavelength. Photon reabsorption and emission to a
longer wavelength could not be realized due to the short
length of EDF used. At 12 m length of EDF, the model
shows that the laser exhibits a relatively high output power
when tuned at 1575 nm.
4. CONCLUSION
We have demonstrated a tuning range of more than 50 nm
from an all-ber ring erbium-doped ber laser with an SNR
of 70 dB. This tunable laser source can be used for wave-
length-division multiplexed systems. A numerical model was
briey discussed and used to compare measured and calcu-
lated results. The experimental results have good matching
values with the modeled results.
REFERENCES
1. Y.T. Chieng, G.J. Cowle, and R.A. Minasian, Optimization of
wavelength tuning of erbium-doped ber ring lasers, J Lightwave
Technol 14 1991., 17301739.
2. A. Frenkel and C. Lin, Inline tunable etalon lter for optical
channel selection in high density wavelength division multiplexed
bre systems, Electron Lett 24 1988., 159161.
3. C.Y. Chen, M.M. Choy, M.J. Andrejco, M.A. Sai and C. Lin, A
widely tunable erbium-doped ber laser pumped at 532 nm, IEEE
Photon Technol Lett 2 1990., 1820.
4. J.L. Zyskind, J.W. Sulhoff, J. Stone, D.J. DiGiovanni, L.W. Stulz,
H.M. Presby, A. Piccirilli, and P.E. Pramayon, Electrically tunable,
diode-pumped erbium-doped bre ring laser with bre Fabry-Perot
etalon, Electron Lett 27 1991., 19501951.
5. H. Schmuck, Th. Pfeiffer, and G. Veith, Widely tunable narrow
linewidth erbium doped bre ring laser, Electron Lett 27 1991.,
21172119.
6. B. Pederson, A. Bjarklev, J.H. Povlsen, K. Dybdal, and C.C.
Larsen, The design of erbium-doped ber ampliers, J Lightwave
Technol 9 1991., 11051112.
7. S. Selvakennedy, M.A. Mahdi, M.K. Abdullah, P. Poopalan, and
H. Ahmad, Design optimisation of erbium-doped bre ring laser
through numerical simulation, Opt Commun 170 1999., 247253.
2001 John Wiley & Sons, Inc.
HILBERT CURVE FRACTAL ANTENNA:
A SMALL RESONANT ANTENNA FOR
VHF / UHF APPLICATIONS
K. J. Vinoy,1 K. A. Jose,1 V. K. Varadan,1 and V. V. Varadan1
1
Center for the Engineering of Electronic and
Acoustic Materials and Devices
The Pennsylvania State University
University Park, Pennsylvania 16802
Recei ed 28 No ember 2000
ABSTRACT: The usefulness of fractal Hilbert cur es in antenna geome-
try is explored here for the rst time. Apart from being simple and
self-similar, these cur es ha e the additional property of approximately
lling a plane. These properties are exploited in realizing a small
resonant antenna. This approach has resulted in an antenna size smaller
than r 10 and still resonant, with performance comparable to a dipole
215
whose resonant length is close to r 2. Numerical predictions of the
input impedance of the antenna ha e been compared with experiments.
The effect of additional fractal iterations on the reduction of the resonant
frequency has been studied. The radiation characteristics of the antenna
at the resonant frequencies pro ided show that this is ery similar to the
dipole characteristics. 2001 John Wiley & Sons, Inc. Microwave Opt
Technol Lett 29: 215219, 2001.
Key words: fractal antennas; Hilbert cur es; multiband antennas;
small antennas
INTRODUCTION
With the widespread proliferation of telecommunication
technology in recent years, the need for small-size multiband
antennas has increased manifold. However, an arbitrary re-
duction in the antenna size would result in a large reactance
and deterioration in the radiation efciency. Meander-line
and zig-zag antennas have been studied for their capability in
antenna size reduction w1, 2x. However, novel antenna con-
gurations using fractal Hilbert curves introduced here can
reduce the antenna size further. Once optimized for radiation
characteristics, these antennas can nd many applications in
UHFrVHF communication antennas.
Two important properties of fractal patterns are self-simi-
larity and scale invariance w3x. Fractals consist of identical or
similar elements repeated in different magnications, orien-
tations, and positions, most often in an interconnected fash-
ion to obtain the nal structure. The seemingly random
nature of fractals is exploited in several elds of engineering
and science. Jaggard and Spielman w4x have shown that the
geometric similarity of fractals can be translated to their
electromagnetic behavior, in the context of diffraction stud-
ies. Similarly, the diffracted elds of self-similar fractal
screens are also found to be self-similar w5x. Although a large
number of fractal patterns have been studied in mathematics,
only very of few of these have actually been implemented for
antenna structures. Among those currently reported in the
literature include Koch curves and the Sierpinski gasket w6,
7x. Some of these geometries have recently been pursued for
antenna applications because of their inherent multiband
nature. However, incorporation of many new fractal geome-
tries into the antenna structures, and various aspects of their
optimization, are still in the incipient stages. Nonetheless, the
analysis of fractal distribution of elements in antenna arrays
has been extensively studied w8x.
The fractal antennas using a Siepinski gasket have been
congured to obtain multiple-frequency bands w9x. These an-
tennas resonate at frequencies in a near-logarithmic interval.
The individual bands at these resonant frequencies are gen-
erally small. However, their relative positions can be con-
trolled by perturbing the shapes of the fractal geometry of
the antenna conguration w10x. This approach has resulted in
a multiband antenna with the individual bands located almost
arbitrarily. Apart from this, other important features of this
fractal antenna include low prole, the possibility of multiple
frequency bands, and moderate gain. The gain of this an-
tenna is quite promising, considering its overall size. These
have been extended further to make them wideband and
conformal w11x.
In this paper, we present a new set of fractal patterns
which are used here in antenna design for the rst time.
These consist of Hilbert curve patterns, which have several
important characteristics hitherto unexplored in antenna en-
gineering.
216 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 29, No. 4, May 20 2001
FRACTAL HILBERT CURVES
Various iteration stages of fractal Hilbert curves are shown in
Figure 1. It may be observed that geometry at a stage can be
obtained by putting together four copies of the previous
iteration, connected to additional line segments. For exam-
ple, the geometry of order 2 can be thought of as four copies
of the geometry with order 1 arranged in different orienta-
tions., connected to the additional segments shown with
dashed lines.
It would be interesting to identify the fractal properties of
this geometry. The plane-lling nature is evident by compar-
ing the rst few iterations of the geometry shown in Figure 1.
It may, however, be mentioned that this geometry is not
strictly self-similar since additional connection segments are
required when an extra iteration order is added to an existing
one. But the contribution of this additional length shown
with dashed lines in Fig. 1. is small compared to the overall
length of the geometry, especially when the order of the
iteration is large. Hence, this small additional length can be
disregarded, which makes the geometry self-similar.
A similar ambiguity also prevails in determining the di-
mension of the geometry. The topological dimension of the
curve is 1 since it consists only of line segments. However, the
dimension of a fractal curve can be dened in terms of a
multiple-copy algorithm w12x. The similarity dimension D is
dened as
log N .
Ds
log 1rf .
where N is the number of copies and f is the scale factor of
consecutive iterations. The dimension of the Hilbert curve is
log w 4 n y 1 . r 4 ny1 y 1 .x
Ds
log w 2 n y 1 . r 2 ny1 y 1 .x
log 4 nr4 ny1 . log 4
f for large n . s s 2.
log 2 r2
n ny1 .
log 2
The similarity dimension of this curve approaches an integer
value 2. because of the approximation involved when a large
fractal order is considerd. But if we consider the length and
number of line segments rst and second iterations, the
dimension is
log 15r3.
D2 s s 1.465.
log 7r3.
The corresponding numbers in the new two iterations are
1.694 and 1.834. These numbers point to the fact that the
dimension of the geometry is still a fractional number, albeit
approaching 2.
Figure 1 Generation of four iterations of Hilbert curves. The
segments used to connect the geometry of the previous iteration are
shown in dashed lines
 As the dimension approaches 2, the curve is almost lling
a plane. In other words, the total length of the line segments
with topological dimension 1. tends to be extremely large.
This could lead to a signicant advantage in antennas since
the resonant frequency can be reduced considerably for a
given area by increasing the fractal iteration order. It may be
recalled that the dimension of this curve is larger than that of
Koch curves dimension s 1.262. studied elsewhere w8x, re-
sulting in a larger reduction factor for the antenna size. The
studies presented here indicate that, by increasing the fractal
iteration order, the resonant frequency of the antenna can be
signicantly reduced. Thus, this approach strives to overcome
one of the fundamental limitations of antenna engineering
with regard to small antennas w13x. It may, however, be
notioned that, since fractals do not come under the purview
of Euclidean geometry, stipulations based on this may be
relaxed for fractals w11x.

NUMERICAL SIMULATIONS
Numerical simulations were done using NEC, which is mo-
ment-method-based software. This is very effective in analyz-
ing antennas that can be modeled with wire segments, such as
the one under consideration here. The model gives accurate
results when the segment length - r20. To suit the re-
quirements, the antenna is modeled without any dielectric
present, although some of the practical implementations do
require dielectric support. A typical antenna geometry with a
third iteration fractal curve is shown in Figure 2. This geome-
try is created with a recursive algorithm consistent with
common fractal generation approaches. Since a reduction in
the overall size of the antenna is of primary importance, this
algorithm is implemented in such a way as to divide this into
equal-sized segments. Thus, for a fractal geometry of third
iteration in Figure 2 nominal outer dimension s 7 cm =
7 cm., each line segment is of length 1 cm.
The feed source point is placed at the point of symmetry
for these structures. The real and imaginary parts of the
input impedance of this Hilbert curve fractal antenna with
two, three, and four iterations are shown in Figure 3. The
antennas occupy a square of side 2, 4, and 8 cm in these
cases. Thus, an antenna with a third iteration fractal consists
of four subgeometries occupying the same area as with the
second iteration considered here. This shows the multiple
resonance characteristics of the antenna, as well as the self-
similarity of its characteristics.
To explore the extent of the plane-lling characteristics of
Figure 2 Antenna conguration with Hilbert curve fractal patterns,
used in simulation
Figure 3 Real and imaginary parts of the input impedance for
three iterations of Hilbert curve fractal antennas obtained with
numerical simulations. The second resonant frequency for the third
iterated curve is slightly lower than the resonant frequency for the
second iterated curve. This may be due to the loading effects, in
addition to the small connector segment added in the fractal genera-
tion. The imaginary part of the fourth iterated antenna shows many
more resonances, and is truncated for clarity
the antenna geometry, these three iterations are also made
within a uniform area of 7 cm = 7 cm, and the corresponding
results are shown in Figure 4. These indicate the self-similar-
ity in the antenna characteristics. Another important charac-
teristic to be noticed is the lowest resonant frequency in each
case. The resonant frequencies are for the third iteration:
360, 980, and 1440 MHz, and for the fourth iteration: 270,
720, 1000, and 1370 MHz. This shows a reduction in the
resonant frequency in this case by 25%. with an increase in
the fractal iteration order, despite the outer dimensions of
the antenna being the same. The predicted radiation patterns
at the resonant frequencies for the third iterated fractal
antenna at its resonant frequencies are shown in Figure 5.
The geometry is placed in the xy-plane, and the pattern cuts
are provided for all three orthogonal planes. These indicate
that, at least for the rst two resonances, the shape of the
radiation pattern remains the same. This is in contrast to a
normal dipole antenna where additional nulls appear with
each subsequent resonances. The reason for this difference is
that the overall size of this radiator remain considerably less
than that of an equivalent linear dipole at these frequencies.

MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 29, No. 4, May 20 2001 217
Figure 4 Computed input impedances for a. third and b. fourth
iteration Hilbert curve antenna. Both antennas have outer dimen-
sions of 7 cm, and are modeled with a wire of 1.3 mm diameter

EXPERIMENTAL RESULTS
The NEC prediction for the third iteration antenna geometry
has been compared with experimental results. The measured
results for this antenna made with copper strips of 4 mm
width are shown in Figure 6. The corresponding simulated
results are presented in Figure 4a.. In the simulation, how-
ever, an equivalent wire diameter is used w14x. The slight
mismatch in this is attributed to the feed arrangement being
unbalanced. The low values for the real part of the impedance
are consistent with other similar small antennas, such as
Koch antennas w15x and small meander-line antennas w2x.
However, by using impedance-matching circuits, or even by
changing the feed location, this can be remedied.

CONCLUSIONS
The effect of self-similarity and plane-lling properties of
fractal Hilbert curves in antenna characteristics is studied
numerically using a moment-method-based software, and the Figure 5 Predicted radiation pattern in two orthogonal planes
predicted input characteristics are compared with experimen- intersecting the antenna geometry, at different resonant frequencies.
tal data. The results presented here establish the link be- The antenna geometry is in the xy-plane. The pattern cuts in the
a. xz-plane, b. yz-plane, and c. xy-plane are shown. The input
tween the self-similarity of the antenna geometries and its
impedance of the geometry used here is shown in Figure 4a.
frequency response. Another important advantage of using
Hilbert curves is the incorporation of its plane-lling charac-
teristics to realize resonant antennas with a smaller overall
physical size. A fourth iterated fractal Hilbert curve geometry
inscribed in a square of side 7 cm is shown to result in a
resonant frequency of 267 MHz, which is much lower than
any other resonant antenna of similar size.

218 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 29, No. 4, May 20 2001
 Figure 6 Measured input impedance of a Hilbert curve fractal antenna. The antenna is fabricated with 4 mm wide copper strips of
the third iterated fractal geometry occupying a 7 cm square area. The measured performances of this antenna is comparable with the
simulated response in Figure 4a.
REFERENCES
1. H. Nakano, H. Tagami, A. Yoshizawa, and J. Yamauchi, Shorten-
ing ratios of modied dipole antennas, IEEE Trans Antennas
Propagat AP-32 1984., 385386.
2. T.J. Warnagiris and T.J. Minardo, Performance of a meandered
line as an electrically small transmitting antenna, IEEE Trans
Antennas Propagat 46 1998., 17971801.
3. B.B. Mandelbrot, The fractal geometry of nature, Freeman, New
York, 1983.
4. D.L. Jaggard and T. Spielman, Triadic cantor target diffraction,
Microwave Opt Technol Lett 5 1992., 460466.
5. A. Lakhtakia, N.S. Holter, V.K. Varadan, and V.V. Varadan,
Self-similarity in diffraction by a self similar fractal screen, IEEE
Trans Antennas Propagat AP-35 1987., 236239.
6. D.H. Werner, R.L. Haupt, and P.L. Werner, Fractal antenna
engineering: The theory and design of fractal antenna arrays,
IEEE Antennas Propagat Mag 41 1999., 3759.
7. C. Puente-Baliarda, J. Romeu, R. Pous, and A. Cardama, On the
behavior of the Sierpinski multiband fractal antenna, IEEE Trans
Antennas Propagat 46 1998., 517524.
8. N. Cohen, Fractal antenna applications in wireless telecommuni-
cations, Prof Program Proc Electron Ind Forum, Boston, MA,
May 1997, pp. 4349.
9. C. Puente-Baliarda, M. Navarro, J. Romeu, and R. Pous, Varia-
tions on the fractal Sierpinski antenna are angle, IEEE AP-S
Int Symp Dig, 1998, pp. 23402344.
10. V.K. Varadan, K. Vinoy, J.A. Kollakompil, and V.V. Varadan,
Miniaturized conformal octave-bandwidth fractal antennas on
high dielectric substrates and chiral layers, U.S. Patent applica-
tion pending, July 2000.
11. C. Puente, J. Romeu, and A. Carsama, Fractal shaped antennas,
Frontiers in electromagnetics, D.H. Werner and R. Mittra Edi-
tors., IEEE Press, New York, 1999, chap. 2.
12. H.O. Peitgen, H. Jurgens, and D. Saupe, Chaos and fractals: New
frontiers of science, Springer-Verlag, New York, 1992.
13. R.C. Hansen, Fundamental limitations in antennas, Proc IEEE
69 1981., 170182.
14. C.A. Balanis, Antenna theory: Analysis and design, Wiley, New
York, 1997, 2nd ed.
MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 29, No. 4, May 20 2001
15. C. Puente-Baliarda, J. Romeu, R. Pous, J. Ramis, and A. Hijazo,
Small but long Koch fractal monopole, Electron Lett 34 1998.,
910.
2001 John Wiley & Sons, Inc.
THREE-PORT RECTANGULAR
MICROSTRIP UNEQUAL POWER
DIVIDER AND COUPLER
Sudhabindu Ray1 and Girish Kumar1
1
Department of Electrical Engineering
Indian Institute of Technology, Bombay
Powai, Mumbai 400 076, India
Recei ed 16 No ember 2000
ABSTRACT: A new conguration of a three-port rectangular microstrip
power di ider and coupler has been proposed. It can easily implement
unequal power di ision of a large ratio, i.e., the power coupled to the
third port can ary from y3 to y26 dB. The conguration has been
analyzed using a transmission-line model, as well as method-of-
moments-based software. 2001 John Wiley & Sons, Inc. Microwave
Opt Technol Lett 29: 219223, 2001.
Key words: rectangular microstrip power di ider; coupler
1. INTRODUCTION
Power dividers and couplers are the building blocks of mi-
crowave and millimeter-wave systems. In a planar congura-
tion, the branch-line directional coupler was rst proposed by
Reed and Wheeler in 1956 w1x, and the n-way power divider
was reported by Wilkinson in 1960 w2x. After that, several
types of power dividers and couplers, including direct-coupled
and coupled-line congurations, were reported w311x. An
unequal power divider can be realized by using a two-way
219